📅 May 12, 2025
📍 Honululu 2 co-located with NOMS 2025
14:00 | Workshop Opening
14:05 | Keynote
🔹 “Intent-based Management in Industry 5.0”
📢 Pal Varga
Abstract
Industry 5.0 is reshaping production and manufacturing paradigms by emphasizing human-centric processes, sustainability, and resilience, taking intelligent automation as a base. A key enabler in realizing this vision is intent-based management – an approach originally conceived for network operations that is increasingly proving transformative in various domains of networking. In this keynote, besides introducing the fundamental concepts of intent-based management, we will outline its potential capabilities throughout the engineering process and discuss use-cases for Industry 5.0. Further, we will demonstrate how explicitly defined human or business intents can streamline the management of complex industrial systems and cyber-physical infrastructures.
In the new operational paradigm of intent-based management, users define high-level business or operational goals – called intents – in a human-friendly manner, abstracting away underlying technical complexities. The system automatically translates these intents into specific configurations, actions, and policies, continuously verifying whether the managed system aligns with the specified objectives. At its core, intent-based management leverages automation, artificial intelligence, and real-time analytics to provide adaptive decision-making, enabling self-configuration, self-monitoring, and self-healing capabilities. The key strength of intent-based management lies in its ability to simplify complex systems management, enhance agility, reduce human error, and swiftly adapt to changing business and environmental conditions. While initially rooted in network management, this approach can extend effectively to various domains, including cyber-physical systems, industrial automation, and infrastructure management.
Components of the intent-based management process stretches from intent specification and translation to automated monitoring and adaptive decision-making. As walking through these process steps, we show how intent-based management simplifies and enhances operational efficiency, responsiveness, and eventually sustainability as well. Additionally, we present real-world use-cases and innovative applications where IBM has empowered dynamic industrial operations, ensuring systems continuously align with strategic business objectives, environmental constraints, and human-centric values inherent in Industry 5.0. Finally, we will discuss critical challenges and future research directions, including real-time intent execution, multi-stakeholder intent resolution, security, explainability, and regulatory considerations.
About Pal Varga
Pal Varga is the Head of Department of Telecommunications and Artificial Intelligence at the Budapest University of Technology and Economics, where he is member of the Senate. His main research interests include communication systems, Cyber-Physical Systems and Industrial IoT, digital twins, network traffic analysis, end-to-end QoS and SLA issues — for which he is keen to apply hardware acceleration and AI/ML techniques as well. He has co-founded the Workshops on Generative AI for Network Management (GAIN at IEEE/IFIP NOMS) and the Workshops on Analytics for Service and Application Management (AnServApp at IEEE/IFIP CNSM). Besides being a member of HTE, he is a senior member of IEEE (both in ComSoc and IES). He is Editorial Board member in many journals, Associate Editor in IEEE TNSM, and the Editor-in-Chief of the Infocommunications Journal.
About Budapest University of Technology and Economics (BME)
Budapest University of Technology and Economics (BME) is one of the oldest and most prestigious institutions of higher education in engineering in Central Europe. Founded in 1782, BME has played an important role in shaping the region’s scientific and technological landscape. The university offers a wide range of programs in engineering, natural sciences, economics, and informatics, and is known for its strong emphasis on research, innovation, and international collaboration.
BME-TMIT, the Department of Telecommunications and Artificial Intelligence at BME brings together expertise in communication systems, signal processing, network engineering, and AI/ML. The department conducts cutting-edge research in 5G/6G networks, cybersecurity, blockchain technologies, IoT systems, speech technology, and AI-driven solutions for intelligent systems and services. It is actively involved in national and European research projects and collaborates closely with industry partners.
14:50 | Technical Session 1
📝 CORECONF Implementation as SDN Southbound Interface for IoT: an OSCORE/EDHOC Use Case
👤 Javier Alejandro Fernandez (IMT Atlantique, France)
👤 Rafa Marin-Lopez (University of Murcia, Spain)
👤 Gabriel Lopez-Millan (University of Murcia, Spain)
👤 Laurent Toutain (IMT Atlantique / IRISA, France)
Abstract. The Internet of Things (IoT) aims to gather valuable data from our surroundings through resource-constrained networks and devices. For this reason, efficient and lightweight communication protocols are required to be developed and adopted. CORECONF, a network management protocol designed for constrained environments, provides a promising solution for IoT device configuration. This work introduces pycoreconf, an open-source implementation of CORECONF, with the goal of testing the protocol and making it more accessible to researchers and developers by enabling its use in real-world scenarios and experimental setups. In this paper, we evaluate its performance and applicability as a southbound interface in an SDN-based architecture, demonstrating its potential for configuring security contexts between IoT devices. Potential for other use cases remains to be explored in future work. Our results suggest that pycoreconf is a viable tool for those interested in exploring and adopting CORECONF in IoT scenarios.
15:10 | Technical Session 2
📝 IoT Connectivity Management by Hyperbolic Trees
👤 Andras Majdan (BME-TMIT, Hungary)
👤 Lejla Pasic (Budapest University of Technology and Economics, Hungary)
👤 Daniel Ficzere (Budapest University of Technology and Economics, Hungary)
👤 Gergely Hollósi (Budapesti Muszaki Egyetem, Hungary)
👤 Heszberger Zalan (Assocp, Hungary)
👤 Rolland Vida (Budapest University of Technology and Economics, Bulgaria)
👤 Jozsef Biro (BME-TMIT, Hungary)
Abstract. The growing scale and complexity of Internet of Things (IoT) networks present critical challenges for maintaining and recovering connectivity, especially in dynamic environments or after large-scale fragmentation caused by disasters. Traditional centralized approaches often face scalability, resilience, and energy efficiency issues. Hyperbolic geometry, which has successfully modeled hierarchical and scale-free structures in large-scale systems such as the Internet’s AS topology, offers a promising foundation for addressing these challenges. This paper introduces hyperbolic trees as a novel framework for IoT connectivity management. Nodes are embedded in a hyperbolic plane, and connectivity is maintained or restored through a simple local rule: each node connects to the nearest node with a smaller radial coordinate. This distributed, asynchronous approach allows nodes to autonomously reorganize and recover connectivity without centralized coordination, ensuring scalability and adaptability. We demonstrate that hyperbolic trees provide a robust, energy-efficient solution for both maintaining and recovering IoT network connectivity. Numerical results validate the framework’s scalability and effectiveness, highlighting its applicability in scenarios ranging from routine management to disaster recovery. These findings establish hyperbolic trees as a practical and scalable tool for resilient IoT connectivity management in next-generation networks.
☕ 15:30 – 16:00 | Coffee Break
16:00 | Technical Session 3
📝 Supervised Learning for Optimal Latency-aware Microservice Placement in Edge Computing Environments
👤 Hirotaka Kasahara (SoftBank Corp., Japan)
👤 Ryuichi Kitajima (SoftBank Corp., Japan)
👤 Jun Towada (SoftBank Corp., Japan)
👤 Osamu Sato (SoftBank Corp., Japan)
Abstract. Edge computing (EC), which facilitates real-time data processing by computing near data source points, is crucial for latency sensitive applications requiring real-time and seamless data processing. A significant challenge in EC is optimal placement applications to ensure that data processing is completed within a specified end-to-end (E2E) timeframe. However, existing container orchestration tools cannot satisfy the E2E latency requirements considering communication latency between computational nodes and processing time in each computational node. This paper proposes a supervised learning-based method for optimal placement of applications in EC environments, and the model’s performance is compared with exact solutions. The results demonstrate that the proposed method learns the optimal placement policy on the microservice-based application. The proposed model surpasses the performance of conventional MLP, GAT, and GCN-based models, achieving a multi-class accuracy of 94.8% on a validation dataset. The difference between the inferred and exact solutions is less than 1.6% compared in the mean objective function. The findings of this study illustrate the effectiveness of the proposed learning-based method for microservice-based applications.
16:20 | Technical Session 4
📝 Exploring Composable Network Stacks from Isolated Components with WebAssembly and QUIC
👤 Benedikt Spies (TUM, Germany)
👤 Christian Obermaier (Technical University of Munich, Germany)
👤 Jörg Ott (Technical University of Munich, Germany)
Abstract. Modern web applications demand increasingly sophisticated network protocols. Deploying applications that integrate custom network algorithms is particularly challenging in container or shared kernel environments, where untrusted code execution must be restricted. While the host-provided network stack is usually highly efficient, it often lacks adaptability. With QUIC emerging as an extensible foundation for network applications and WebAssembly (WASM) offering a secure, lightweight container runtime, we propose a modular network stack architecture composable of host and isolated guest components. This paper investigates the performance cost of implementing a complete network stack—including HTTP/3, QUIC, and TLS— as isolated WASM components. We evaluate our architecture, built on the WebAssembly System Interface (WASI) 0.2 and the WASM Component Model, through a comprehensive comparison with a native implementation. Our results demonstrate that while building a modular QUIC stack with WASM is technically feasible, its current performance overhead limits its practical use. The composed WASM implementation achieves a maximum goodput of 0.9 Gbps, compared to 6.1 Gbps with a native implementation. We identified the lack of send-offloading capabilities, missing cryptographic acceleration, and additional memory copies as the mayor performance bottlenecks. We discuss these limitations and propose potential improvements to bridge the performance gap.
16:40 | Technical Session 5
📝 Performance Observations from Split Federated Learning on Heterogeneous Devices and Networks
👤 Samuel Trepac (Mount Royal University, Canada)
👤 Yasaman Amannejad (Mount Royal University, Canada)
Abstract. Split Federated Learning (SFL) combines the scalability of Federated Learning (FL) with the computational efficiency of Split Learning (SL), making it a promising paradigm for distributed machine learning in heterogeneous environments. While existing work have explored theoretical convergence and split layer selection, the practical implications of split layer selection with client heterogeneity and network variability remain under explored. This paper investigates the impact of heterogeneous client resources, network conditions, and split layer configurations on the performance of SFL. Using the CIFAR-10 dataset, we implement a SFL model with heterogeneous clients. Experiments include diverse configurations, such as uniform and varying split layers among clients. We analyze training time, idle and active times, and resource utilization to understand the effects of heterogeneity. Our observations show that strategic split layer configurations tailored to heterogeneous environments can improve training efficiency by 26.7%. Additionally we observed the effects that heterogeneous networks have on SFL and how to compensate for it, offering valuable insights for deploying SFL in real-world computing systems.
17:00 | Technical Session 6
📝 Handling Multiple Events in IoT Environments: A Workflow-Based FaaS System
👤 Jangwon Seo (SoongSil University, South Korea)
👤 Hyeongwoo Ju (SoongSil University, South Korea)
👤 Younghan Kim (SoongSil University, South Korea)
Abstract. IoT (Internet of Things) systems are increasingly implemented as Function as a Service (FaaS) on edge cloud platforms, enabling efficient utilization of edge cloud resources by executing services based on various events. Recently, research has been focused on extending event-driven FaaS frameworks to support complex event processing, meeting the diverse requirements of modern IoT applications. However, implementing multi-event processing based on predefined scenarios and fixed functions introduces limitations in accommodating new event types, integrating new functions, modifying existing functions, and dynamically reconfiguring event-function mappings. These constraints hinder the operational flexibility and scalability required in dynamic IoT environments. To address these challenges, this paper proposes a workflow-based FaaS framework that enables flexible function modification and extension while supporting the reuse of existing functions. The proposed system is implemented to demonstrate the efficiency of system operation by complementing the limitations of the monolithic implementation method of existing complex event processing systems and is evaluated in terms of response time and performance. The evaluation results showed that compared to the existing fixed scenario-based monolithic event processing system, our system has flexibility in system operation, scalability, reusability of functions, and operates within an acceptable processing time at the IoT service level.